摘要
本文研究了基于互信息、相关性的特征选择方法,并介入网页页面中超链接因素,对特征提取中互信息计算公式作了改进-引入超链接因子。实验表明,改进之后比之以往的简单的基于互信息方法进行特征选择的网页分类精度有一定的提高。
This paper is a study on feature selection based on MI and relativity.In view of hyperlinks of web pages,we improve the algorithem of MI in selection of feature in which factor of hyperlink is inducted, Experiment shows that,it is more accurate that web pages are classificated in this new method than ever.
出处
《科技广场》
2009年第9期39-40,共2页
Science Mosaic
关键词
网页分类
特征提取
互信息
Web Page Classification
Feature Selection
MI
